Introduction

Importing libraries

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import plotly.express as px
import seaborn as sns

from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split

from sklearn.naive_bayes import BernoulliNB 
from sklearn.svm import LinearSVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier

from sklearn.preprocessing import LabelBinarizer

from sklearn.metrics import accuracy_score, confusion_matrix, precision_score, f1_score, recall_score, roc_curve, roc_auc_score

Introduction

Objective: The Objective of the project is to accurately predict the chances of developing lung cancer based off the genetic risk target variable.

Background: Cancer is a disease in which cells in the body grow out of control. When cancer starts in the lungs, it is called lung cancer.

Lung cancer begins in the lungs and may spread to lymph nodes or other organs in the body, such as the brain. Cancer from other organs also may spread to the lungs. When cancer cells spread from one organ to another, they are called metastases.

Lung cancers usually are grouped into two main types called small cell and non-small cell (including adenocarcinoma and squamous cell carcinoma). These types of lung cancer grow differently and are treated differently. Non-small cell lung cancer is more common than small cell lung cancer.

About the dataset

Reading the dataset

dataset = pd.read_csv('/Users/abdihussein/Downloads/cancer patient data sets.csv', sep = ',')
dataset.head()

	index	Patient Id	Age	Gender	Air Pollution	Alcohol use	Dust Allergy	OccuPational Hazards	Genetic Risk	chronic Lung Disease	...	Fatigue	Weight Loss	Shortness of Breath	Wheezing	Swallowing Difficulty	Clubbing of Finger Nails	Frequent Cold	Dry Cough	Snoring	Level
0	0	P1	33	1	2	4	5	4	3	2	...	3	4	2	2	3	1	2	3	4	Low
1	1	P10	17	1	3	1	5	3	4	2	...	1	3	7	8	6	2	1	7	2	Medium
2	2	P100	35	1	4	5	6	5	5	4	...	8	7	9	2	1	4	6	7	2	High
3	3	P1000	37	1	7	7	7	7	6	7	...	4	2	3	1	4	5	6	7	5	High
4	4	P101	46	1	6	8	7	7	7	6	...	3	2	4	1	4	2	4	2	3	High

5 rows × 26 columns

dataset.isnull().sum()

index                       0
Patient Id                  0
Age                         0
Gender                      0
Air Pollution               0
Alcohol use                 0
Dust Allergy                0
OccuPational Hazards        0
Genetic Risk                0
chronic Lung Disease        0
Balanced Diet               0
Obesity                     0
Smoking                     0
Passive Smoker              0
Chest Pain                  0
Coughing of Blood           0
Fatigue                     0
Weight Loss                 0
Shortness of Breath         0
Wheezing                    0
Swallowing Difficulty       0
Clubbing of Finger Nails    0
Frequent Cold               0
Dry Cough                   0
Snoring                     0
Level                       0
dtype: int64

dataset = dataset.iloc[:, 2:]

dataset.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 24 columns):
 #   Column                    Non-Null Count  Dtype 
---  ------                    --------------  ----- 
 0   Age                       1000 non-null   int64 
 1   Gender                    1000 non-null   int64 
 2   Air Pollution             1000 non-null   int64 
 3   Alcohol use               1000 non-null   int64 
 4   Dust Allergy              1000 non-null   int64 
 5   OccuPational Hazards      1000 non-null   int64 
 6   Genetic Risk              1000 non-null   int64 
 7   chronic Lung Disease      1000 non-null   int64 
 8   Balanced Diet             1000 non-null   int64 
 9   Obesity                   1000 non-null   int64 
 10  Smoking                   1000 non-null   int64 
 11  Passive Smoker            1000 non-null   int64 
 12  Chest Pain                1000 non-null   int64 
 13  Coughing of Blood         1000 non-null   int64 
 14  Fatigue                   1000 non-null   int64 
 15  Weight Loss               1000 non-null   int64 
 16  Shortness of Breath       1000 non-null   int64 
 17  Wheezing                  1000 non-null   int64 
 18  Swallowing Difficulty     1000 non-null   int64 
 19  Clubbing of Finger Nails  1000 non-null   int64 
 20  Frequent Cold             1000 non-null   int64 
 21  Dry Cough                 1000 non-null   int64 
 22  Snoring                   1000 non-null   int64 
 23  Level                     1000 non-null   object
dtypes: int64(23), object(1)
memory usage: 187.6+ KB

a=dataset.hist(figsize=(15,15))

Showing the informations

fig = px.histogram(dataset, x = 'Age', color = 'Level', nbins = 10, title = 'Distribution of ages', text_auto = True)
fig.show()

fig = px.box(dataset, x = 'Age', y = 'Level')
fig.show()

plt.figure(figsize = (6,4))
ax = sns.countplot(dataset.Level)
for bars in ax.containers:
    ax.bar_label(bars)
plt.title("Count of Levels", fontsize = 15);

def graph(x, y, title):
  fig = px.histogram(dataset, x = x, y = y, color = 'Level', barmode = 'group', 
                   title = title, text_auto = True)
  fig.show()

  fig = px.box(dataset, x = x, y = 'Level')
  fig.show()

#graph('Alcohol use', 'Gender', 'Distribution of Alcohol use')

graph('Genetic Risk', 'Gender', 'Distribution of Genetic Risk')

fig = px.histogram(dataset, x = 'Level', y = 'Gender', color = 'Gender', barmode = 'group', 
                   title = 'Distribution of Level', text_auto = True)
fig.show()

What the data shows for us?

It's possible to understand that according the level of column is increasing, higher is the level of people risk. There are few examples where exist some outliers. Perhaps, the medium risk isn't so objective, because it's the average between high and low values, and it isn't such a clear category to define specifically.

dataset.describe()

	Age	Gender	Air Pollution	Alcohol use	Dust Allergy	OccuPational Hazards	Genetic Risk	chronic Lung Disease	Balanced Diet	Obesity	...	Coughing of Blood	Fatigue	Weight Loss	Shortness of Breath	Wheezing	Swallowing Difficulty	Clubbing of Finger Nails	Frequent Cold	Dry Cough	Snoring
count	1000.000000	1000.000000	1000.0000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	...	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000	1000.000000
mean	37.174000	1.402000	3.8400	4.563000	5.165000	4.840000	4.580000	4.380000	4.491000	4.465000	...	4.859000	3.856000	3.855000	4.240000	3.777000	3.746000	3.923000	3.536000	3.853000	2.926000
std	12.005493	0.490547	2.0304	2.620477	1.980833	2.107805	2.126999	1.848518	2.135528	2.124921	...	2.427965	2.244616	2.206546	2.285087	2.041921	2.270383	2.388048	1.832502	2.039007	1.474686
min	14.000000	1.000000	1.0000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	...	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000
25%	27.750000	1.000000	2.0000	2.000000	4.000000	3.000000	2.000000	3.000000	2.000000	3.000000	...	3.000000	2.000000	2.000000	2.000000	2.000000	2.000000	2.000000	2.000000	2.000000	2.000000
50%	36.000000	1.000000	3.0000	5.000000	6.000000	5.000000	5.000000	4.000000	4.000000	4.000000	...	4.000000	3.000000	3.000000	4.000000	4.000000	4.000000	4.000000	3.000000	4.000000	3.000000
75%	45.000000	2.000000	6.0000	7.000000	7.000000	7.000000	7.000000	6.000000	7.000000	7.000000	...	7.000000	5.000000	6.000000	6.000000	5.000000	5.000000	5.000000	5.000000	6.000000	4.000000
max	73.000000	2.000000	8.0000	8.000000	8.000000	8.000000	7.000000	7.000000	7.000000	7.000000	...	9.000000	9.000000	8.000000	9.000000	8.000000	8.000000	9.000000	7.000000	7.000000	7.000000

8 rows × 23 columns

dataset

	Age	Gender	Air Pollution	Alcohol use	Dust Allergy	OccuPational Hazards	Genetic Risk	chronic Lung Disease	Balanced Diet	Obesity	...	Fatigue	Weight Loss	Shortness of Breath	Wheezing	Swallowing Difficulty	Clubbing of Finger Nails	Frequent Cold	Dry Cough	Snoring	Level
0	33	1	2	4	5	4	3	2	2	4	...	3	4	2	2	3	1	2	3	4	Low
1	17	1	3	1	5	3	4	2	2	2	...	1	3	7	8	6	2	1	7	2	Medium
2	35	1	4	5	6	5	5	4	6	7	...	8	7	9	2	1	4	6	7	2	High
3	37	1	7	7	7	7	6	7	7	7	...	4	2	3	1	4	5	6	7	5	High
4	46	1	6	8	7	7	7	6	7	7	...	3	2	4	1	4	2	4	2	3	High
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
995	44	1	6	7	7	7	7	6	7	7	...	5	3	2	7	8	2	4	5	3	High
996	37	2	6	8	7	7	7	6	7	7	...	9	6	5	7	2	4	3	1	4	High
997	25	2	4	5	6	5	5	4	6	7	...	8	7	9	2	1	4	6	7	2	High
998	18	2	6	8	7	7	7	6	7	7	...	3	2	4	1	4	2	4	2	3	High
999	47	1	6	5	6	5	5	4	6	7	...	8	7	9	2	1	4	6	7	2	High

1000 rows × 24 columns

Standarding the data

dataset['Level'].replace(to_replace=["Low", "Medium", "High"], value = [1, 2, 3], inplace = True)

standard = StandardScaler()
dataset_standard = standard.fit_transform(dataset)

dataset_standard_2 = pd.DataFrame(data = dataset_standard, columns = dataset.keys())

correlation = dataset_standard_2.corr()
graph = sns.heatmap(correlation, annot = True, fmt = '.2f')
graph.figure.set_size_inches(24, 16)
graph.set_title('Correlation Matriz', fontsize = 30)

Text(0.5, 1.0, 'Correlation Matriz')

The data for train and test

X = dataset_standard_2.iloc[:, :23].values
y = dataset.iloc[:, 23].values

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.25, random_state = 1234)

Running the models

from sklearn.metrics import confusion_matrix

bnb = BernoulliNB()
bnb.fit(X_train, y_train)
bnb_predict = bnb.predict(X_test)

accuracy_bnb = accuracy_score(y_test, bnb_predict)

confusion_matrix = confusion_matrix(y_test, bnb_predict)
plt.figure(figsize = (6,6))
sns.set(font_scale = 1)
sns.heatmap(confusion_matrix, annot = True, fmt = '.1f').set(xlabel = 'Prediction', ylabel = 'Real')

print('Accuracy: {:.2f}'.format(accuracy_bnb))

Accuracy: 0.82

from sklearn.metrics import confusion_matrix

svc = LinearSVC()
svc.fit(X_train, y_train)
svc_predict = svc.predict(X_test)

accuracy_svc = accuracy_score(y_test, svc_predict)

confusion_matrix = confusion_matrix(y_test, svc_predict)
plt.figure(figsize = (6,6))
sns.set(font_scale = 1)
sns.heatmap(confusion_matrix, annot = True, fmt = '.1f').set(xlabel = 'Prediction', ylabel = 'Real')

print('Accuracy: {:.2f}'.format(accuracy_svc))

Accuracy: 1.00

from sklearn.metrics import confusion_matrix

dtc = DecisionTreeClassifier(criterion='entropy', random_state = 1234)
dtc.fit(X_train, y_train)
dtc_predict = dtc.predict(X_test)

accuracy_dtc = accuracy_score(y_test, dtc_predict)

confusion_matrix = confusion_matrix(y_test, dtc_predict)
plt.figure(figsize = (6,6))
sns.set(font_scale = 1)
sns.heatmap(confusion_matrix, annot = True, fmt = '.1f').set(xlabel = 'Prediction', ylabel = 'Real')

print('Accuracy: {:.2f}'.format(accuracy_dtc))

Accuracy: 1.00

from sklearn.metrics import confusion_matrix

rfc = RandomForestClassifier(n_estimators = 100)
rfc.fit(X_train, y_train)
rfc_predict = rfc.predict(X_test)

accuracy_rfc = accuracy_score(y_test, rfc_predict)

confusion_matrix = confusion_matrix(y_test, rfc_predict)
plt.figure(figsize = (6,6))
sns.set(font_scale = 1)
sns.heatmap(confusion_matrix, annot = True, fmt = '.1f').set(xlabel = 'Prediction', ylabel = 'Real')

print('Accuracy: {:.2f}'.format(accuracy_rfc))

Accuracy: 1.00

Calculating the ROC curve of each model

lb = LabelBinarizer()
y_test = lb.fit_transform(y_test)
bnb_predict = lb.fit_transform(bnb_predict)
svc_predict = lb.fit_transform(svc_predict)
knn_predict = lb.fit_transform(knn_predict)
dtc_predict = lb.fit_transform(dtc_predict)
rfc_predict = lb.fit_transform(rfc_predict)

bnb_auc = [0] * 3
bnb_fpr = [0] * 3
bnb_tpr = [0] * 3
bnb_thresholds = [0] * 3

svc_auc = [0] * 3
svc_fpr = [0] * 3
svc_tpr = [0] * 3
svc_thresholds = [0] * 3

knn_auc = [0] * 3
knn_fpr = [0] * 3
knn_tpr = [0] * 3
knn_thresholds = [0] * 3

dtc_auc = [0] * 3
dtc_fpr = [0] * 3
dtc_tpr = [0] * 3
dtc_thresholds = [0] * 3

rfc_auc = [0] * 3
rfc_fpr = [0] * 3
rfc_tpr = [0] * 3
rfc_thresholds = [0] * 3

for i in range(len(bnb_auc)):
  bnb_auc[i] = roc_auc_score(y_test[:, i], bnb_predict[:, i], multi_class = 'ovr', average = 'macro')
  svc_auc[i] = roc_auc_score(y_test[:, i], svc_predict[:, i], multi_class = 'ovr', average = 'macro')
  knn_auc[i] = roc_auc_score(y_test[:, i], knn_predict[:, i], multi_class = 'ovr', average = 'macro')
  dtc_auc[i] = roc_auc_score(y_test[:, i], dtc_predict[:, i], multi_class = 'ovr', average = 'macro')
  rfc_auc[i] = roc_auc_score(y_test[:, i], rfc_predict[:, i], multi_class = 'ovr', average = 'macro')
  bnb_fpr[i], bnb_tpr[i], bnb_thresholds[i] = roc_curve(y_test[:, i], bnb_predict[:, i])
  svc_fpr[i], svc_tpr[i], svc_thresholds[i] = roc_curve(y_test[:, i], svc_predict[:, i])
  knn_fpr[i], knn_tpr[i], knn_thresholds[i] = roc_curve(y_test[:, i], knn_predict[:, i])
  dtc_fpr[i], dtc_tpr[i], dtc_thresholds[i] = roc_curve(y_test[:, i], dtc_predict[:, i])
  rfc_fpr[i], rfc_tpr[i], rfc_thresholds[i] = roc_curve(y_test[:, i], rfc_predict[:, i])

plt.figure(figsize = (7, 7))
plt.plot(dtc_fpr[0], dtc_tpr[0], label = 'Dec. Tree Class. Model (Low) - ROC Curve (area = %.2f)' %dtc_auc[0])
plt.plot(dtc_fpr[1], dtc_tpr[1], label = 'Dec. Tree Class. Model (Medium) - ROC Curve (area = %.2f)' %dtc_auc[1])
plt.plot(dtc_fpr[2], dtc_tpr[2], label = 'Dec. Tree Class. Model (High) - ROC Curve (area = %.2f)' %dtc_auc[2])
plt.plot([0, 1], [0, 1], linestyle = '--', color = 'r', label = 'Random guess')
plt.title('ROC curve (Decision Tree Classifier)')
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.grid(True)
plt.legend()
plt.show()

Conclusion

Except the Naive Bayes, the rest of models have satisfied number of accuracy. However, the test data was small (250) to be sure that these models will be right every time when it insert new informations to predict. It'll be necessary a bigger dataset to train more and to analysis if the accuracy'll be the same or it will be decrease.